Problem 355
Medium Theorems and Indeterminate Forms

Determine which condition of Rolle's theorem fails

Determine which condition of Rolle's theorem fails for $f(x) = \bigl|\sin x\bigr|$ on $\left[-\dfrac{\pi}{2},\dfrac{\pi}{2}\right]$.

Explanation of Method

A corner at the origin, where the one-sided slopes are +1 and −1.

Step 1: Continuity holds

$\sin x$ is continuous and so is the absolute value, so their composition is continuous everywhere.

Step 2: The endpoint values agree

\[ f\left(-\frac{\pi}{2}\right) = \bigl|-1\bigr| = 1, \qquad f\left(\frac{\pi}{2}\right) = \bigl|1\bigr| = 1. \]

Step 3: Where differentiability fails

On the interval $\sin x$ is negative to the left of $0$ and positive to the right, so

\[ f(x) = \begin{cases} -\sin x, & -\tfrac{\pi}{2}\lt x\lt 0,\\ \ \ \sin x, & 0\lt x\lt \tfrac{\pi}{2}, \end{cases} \qquad f'(x) = \begin{cases} -\cos x, & x\lt 0,\\ \ \ \cos x, & x>0. \end{cases} \]

At the origin the difference quotient has different one-sided limits: $\lim_{h\to0^{+}}\frac{\sin h}{h} = +1, \qquad \lim_{h\to0^{-}}\frac{-\sin h}{h} = -1$.

Since $+1\neq-1$, the derivative $f'(0)$ does not exist.

Step 4: The conclusion fails too

For $x\neq0$ in the interval, $f'(x) = \pm\cos x$, and $\cos x\neq0$ there; at $x = 0$ the derivative does not exist. So no $\xi$ satisfies $f'(\xi) = 0$.

Step 5: Verification

\[f'\left(-\frac{\pi}{4}\right) = -\frac{\sqrt2}{2} \approx -0.707, \qquad f'\left(\frac{\pi}{4}\right) = +\frac{\sqrt2}{2} \;\;\checkmark\]

\[\text{Neither vanishes anywhere on the interval, and } \bigl|f'\bigr|\ge\cos\frac{\pi}{2} = 0 \text{ only in the limit} \;\;\checkmark\]

\[\text{Minimum of } f \text{ is } 0 \text{ at } x = 0, \text{ attained at the non-differentiable point} \;\;\checkmark\]


Result

Final answer:

\[\boxed{\,\text{The failed condition is differentiability: } f'(0) \text{ does not exist, } 0\in\left(-\frac{\pi}{2},\frac{\pi}{2}\right)\,}\]

Remark

Remark. The origin is a corner, not a cusp: the one-sided tangents have finite slopes $\mp1$ and simply fail to agree, unlike Problem 354 where they become vertical. Either kind of break is enough to defeat Rolle's theorem, and in both problems the interior extremum sits exactly at the bad point — which is no coincidence, since that is where the proof would have looked for it.